Junie is JetBrains' own AI coding agent. For Junie, Paqad is advisory: onboarding writes .junie/AGENTS.md, telling the AI how to load your project's rules and workflows, but nothing in Junie makes the AI follow it and Paqad installs no hooks for Junie. If you work in a JetBrains IDE and want Paqad's checks to actually run, use the Claude Code plugin instead, as explained below.
What Paqad can enforce in Junie
Nothing, during the session. Junie is at the advisory level: the tool is given the rules but nothing stops it ignoring them.
- No edit is refused by Paqad.
- No verification runs when Junie finishes.
- Paqad's evidence files are only written if the AI runs Paqad's commands itself.
Paqad's source records Junie as advisory, with hooks, skills, agents, cache and memory switched off for this adapter.
Files onboarding writes
| File | What it is |
|---|---|
.junie/AGENTS.md | The entry file. It tells the AI to open .paqad/framework-path.txt and follow the installed framework, or to act as a normal assistant if Paqad is missing or off. An existing file keeps its text; Paqad's lines go in a marked block. |
.junie/mcp/mcp.json | The MCP servers Paqad suggests for your stack. MCP (Model Context Protocol) lets AI tools connect to extra information sources. The file names servers but holds no start commands, so it connects nothing by itself. |
Paqad 1.91.1 does not write .junie/guidelines.md. If your Junie version only reads guidelines.md, ask Junie to read .junie/AGENTS.md at the start of a session.
Three ways to use AI in a JetBrains IDE
"AI in a JetBrains IDE" can mean three different products, and Paqad treats them differently:
| Product | Paqad adapter | Level |
|---|---|---|
| Junie, JetBrains' own agent | junie | Advisory. Reads Paqad's instructions; no hooks. |
| Claude Agent inside JetBrains AI Assistant | aiassistant, which writes .aiassistant/rules/guidelines.md | Advisory. Reads Paqad's instructions; it has no hook mechanism Paqad can use. |
The Claude Code plugin for JetBrains, which runs the real claude command in the IDE's terminal | claude-code | Blocks and checks, exactly as on the Claude Code page. |
If your team needs Paqad's checks to run inside a JetBrains IDE, the Claude Code plugin is the supported route.
Getting evidence without hooks
Paqad's instructions ask the AI, on advisory tools, to record each stage of feature work by running Paqad's commands in the terminal:
npx paqad-ai stage start planning
npx paqad-ai stage end planning --artifact <plan file>These write to the ledger, Paqad's evidence files under .paqad/ledger/. Nothing checks that Junie runs them. An AI that only describes the stages in chat leaves the ledger empty, so look in .paqad/ledger/paqad.stage-evidence/ after a change.
Independently of the AI, git post-commit and post-merge hooks link commits to their feature, and you can add Paqad's verification script to your CI pipeline. See Adapters overview.
Setting it up
- Run
paqad-ai onboard --providers junie, or leave Junie ticked when onboarding asks (it is ticked by default). - Open the project and check that Junie has read
.junie/AGENTS.md.